基于径向基函数神经网络的板形模式识别研究  

Research on the flatness pattern recognition based on radial basic function network

作　　者：张秀玲[1] 陈丽杰[2] 季颖[1] 逄宗鹏[1] 

机构地区：[1]燕山大学电气工程学院,河北秦皇岛066004 [2]燕山大学工业计算机控制工程河北省重点实验室,河北秦皇岛066004

出　　处：《工业仪表与自动化装置》2009年第3期7-9,共3页Industrial Instrumentation & Automation

基　　金：国家自然科学基金资助项目(50675186)

摘　　要：针对板带轧制过程中用于辨识板形模式的网络精度较低、在线速度较慢和获得网络辨识模型较复杂的问题,提出了一种基于径向基函数神经网络(RBF)的板形模式识别方法。该方法使输入节点减少,网络模型简化,并用模糊C均值算法和伪逆法确定RBF网络的参数,解决了传统方法学习时间较长的问题。实验表明,该方法能有效的提高板形模式识别的精度和速度。Flatness pattern recognition based on radial basic function （RBF） network was put forward in view of the feature that the accuracy of recognition network is low , the on - line speed is slow and the obtainment of network identification model is complex. This method basis fuzzy distance reduces the input node and makes the network model simple. And the method uses the fuzzy C - means clustering and the pseudo - inverse to determine the parameter of RBF network, which had solved the problem of the study time in traditional method is long. The experiment has show that this method can enhance the accuracy and speed of flatness pattern recognition effectively.

关 键 词：板形模式识别 RBF网络 模糊C均值算法 伪逆法 

分 类 号：O235[理学—运筹学与控制论]